Normalize pair coalescence counts/rates using non-missing span

[nspope]

Fixes #3053

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 89.86%. Comparing base (b31d4d1) to head (29f2c92).
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3059   +/-   ##
=======================================
  Coverage   89.85%   89.86%           
=======================================
  Files          29       29           
  Lines       32137    32150   +13     
  Branches     5765     5768    +3     
=======================================
+ Hits        28877    28890   +13     
  Misses       1859     1859           
  Partials     1401     1401           

Flag	Coverage Δ
c-tests	86.71% <100.00%> (+0.01%)	⬆️
lwt-tests	80.78% <ø> (ø)
python-c-tests	89.05% <ø> (ø)
python-tests	98.98% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
c/tskit/trees.c	90.70% <100.00%> (+0.02%)	⬆️
python/tskit/trees.py	98.80% <ø> (ø)

---- 🚨 Try these New Features:

• Flaky Tests Detection - Detect and resolve failed and flaky tests

[jeromekelleher]

Nice, great use of the num_edges count 👍

[jeromekelleher]

We do need to update the changelog for this, though, as it's a significant bugfix

[nspope]

What's the best way to check for NAN in the C library? I see there's a tsk_is_unknown_time, but this returns false for tsk_is_unknown_time(NAN) on my system (Ubuntu 20.04 x86_64). Stack Overflow suggests that NAN != NAN is portable (guaranteed by IEEE 754) ... ?

Edit ah whoops I see there's an isnan macro in C99

[nspope]

The changelog is updated, and I added some tests checking the behavior of the statistics on empty windows. In particular, pair_coalescence_counts will return zeros (rather than NaNs) if span_normalise=True and a window has no nonmissing span; the other two functions will return NaNs.

[jeromekelleher]

LGTM. Great to see these awkward corner cases getting hit so soon, this stuff is definitely getting used!